Big Cube BLD Discussion

[LarsN]

I always memo centers by position, precisely because it makes memo significantly faster.
For execution, I don't do cycles. I just solve a whole center, then another one, and so on.

Oh, I didn't know that anybody did this. How do you solve the centers?

 

[Marcell]

Yeah, could you elaborate? The last thing I heard about this was that there is no efficient solving method for this kind of "static" memo.

 

[Zane_C]

I've been doing a little experiment lately. People always say that you should memo piece cycles and not piece position. I decided to test if this was true and found that bigcube centers would be an area which might have some potentiel, because they are not as restricted as other piece types.

Then I developed a memo method (I did not want to do this visually!) and I found that memo was not a problem at all. I didn't suspect it to be, but it seems that memo will be at least as fast or maybe even faster than cycle memo. Here is my memo from my last test solve on 4x4 centers:

"Han Solo attacks Kristoffer (my cousin)"
"Cinderella falls dead on the body of an alien"
"Belle (beauty and the beast) lies dead at the foot af a pilot"

Each person and the persons action defines the colors of the four positions on a face of the cube. Not doing an action is also an action. I place the three images in the same room (in this case my living room) because I need to be able to random access the faces.
Very little information is needed to remember the colors of all 24 pieces. This is why I believe that memo could be faster then cycle memo with practice.

The hard part is of course execution. I've only had two attempts so far, but they where not as slow as I feared. The problems I encountered so far:

- It takes some time to find the cycles to solve. I believe that this time will reduce with practise but not likely to be as fast as prepared cycle memo.
- Remembering which pieces you have already finished. This is a big hurdle for me, but of course I need more practice.

Both my first attempts were successful and I feel very comfortable about the memo part. But I was wondering if anybody had ideas on how to approach execution in a better way. I doubt that this will become my main method, but I would like to test it some more. Maybe it will be useful for a 7x7x7bld because of all the centers.

Interesting read. The execution seems like it would take a lot of thinking, but as you said, practice would reduce the effort.
Even if the memo was somewhat faster than regular cycling memo, the slowed execution could be a major downfall.

 

[cmhardw]

Richard Patterson had the idea to come up with a 1296 image system that would allow you to encode all the pieces in one orbit on the same face using only 1 image. Also, 1296 is not that many images compared to our current lists, usually around the 550's.

I don't personally have any experience with how to deal with the solving phase, but I do like the idea of solving one face at a time. Is there anyway to implement that using 3-cycles rather than 2-cycles? I'm not even sure there would be one buffer location when solving one face at a time, as it wouldn't be very often that the buffer spot was the color you needed to cycle.

Very interesting. If a good method came out for this that was worth switching to, I would definitely begin building a 1296 image system though. I feel if you're going to do this, then go all out.

 

[TMOY]

Yeah, could you elaborate? The last thing I heard about this was that there is no efficient solving method for this kind of "static" memo.

Depends on what you call "efficient". Solving with static memo 'is slower than solving by cycles, but if you lose less time on execution than you're saving on memo it's still efficient enough.

Let's say I'm solving a 4^3 (for bigger cubes it works the same way). I start with a center with 2 or 3 cubies of identical colors (if any, if there's none the scramble sucks anyway) and put the remaining ones on it using niklas, solving some other cener cubies in the process if possible. Then I do the same on a 2nd center, and so on. If the centers are not solved at their correct places, I use whole center swaps at the end to fix them.

The main difficulty of the method is to keep track of everything you're doing to centers, but with practice it's not that hard.

I don't personally have any experience with how to deal with the solving phase, but I do like the idea of solving one face at a time. Is there anyway to implement that using 3-cycles rather than 2-cycles? I'm not even sure there would be one buffer location when solving one face at a time, as it wouldn't be very often that the buffer spot was the color you needed to cycle.

I think it's definitely possible to use 3-cycles, I just haven't practised them yet. For the buffer, I use a floating buffer anyway.

 

[LarsN]

I tried a solve doing a face at a time. It seemed just as difficult as regular 3-cycling. I think I will stick with 3 cycles for a time and try to improve.

Richard Patterson had the idea to come up with a 1296 image system that would allow you to encode all the pieces in one orbit on the same face using only 1 image. Also, 1296 is not that many images compared to our current lists, usually around the 550's.

1296 images, is that for a 3x3x3 (4 corners and 4 edges) or 4x4x4 (4 centers, 4 corners and 8 wings)? I don't really get it.

My system has 441 persons doing 1 of four action types to define 4 centers on a face.

 

[riffz]

I tried a solve doing a face at a time. It seemed just as difficult as regular 3-cycling. I think I will stick with 3 cycles for a time and try to improve.



1296 images, is that for a 3x3x3 (4 corners and 4 edges) or 4x4x4 (4 centers, 4 corners and 8 wings)? I don't really get it.

My system has 441 persons doing 1 of four action types to define 4 centers on a face.

An image for every possible configuration of 4x4 centers on a given face.

6 possible colours per piece, and 4 center pieces per face gives 6^4 = 1296

I haven't heard of any efficient method for solving pieces after memorizing the current state of the cube in this way, but it would be amazing to memo centers using only 6 images.

 

[LarsN]

but it would be amazing to memo centers using only 6 images.

That's what I do with my memo. Using only 441 images/persons. You can check the example from my earlier post or I can explain if anybody is interested.

 

[rock1313]

Theres a problem I always encounter in big cube centers. When I memorize a piece in one face and then later I memorize the piece in the same face but accidentally memorize the same piece I memorized before and at the end of my memorization. I know something is wrong but it takes me ages to find out what is the problem. For example I memorized the letter N and at I know at the end of the memorization I have done something wrong. I go through my memorization and eventually I found out I memorized 2 N's instead of 1.

How can I keep track of center pieces I've already memorized and don't memorize the same piece twice.

 

[Marcell]

Use your fingers: put one finger on the center piece you memorized last on each sides. (It is neccessery that you always start from the same piece and move in the same direction.)

 

[rock1313]

(It is neccessery that you always start from the same piece and move in the same direction.)

I never thought of it that way.

Thanks

 

[MrMoney]

Corners and midges depend of each other. If you have no corner parity, you have no midge parity either. If you do have corner parity, you also have midge parity.
For that I do: Corners + Y-Perm (make sure that the centers don't get affected if you do corners first). Then I do my solve with midges last. After that: U' F2 U m2 U' F2 U, that fixes the m-slice and the UL and UB midge. Then I do y2 R U R' U' (Rw2 F2 U2 r2 U2 F2 Rw2) U R U' R' y2. And after that a Y-Perm is left.

About 5BLD, am I correct in this?:

Can anyone write this even simpler? I am sorry but it still is not very clear to me. Do we do pairs of corners untill last corner, then midges with Ui F2 U m2 Ui F2 U, then y2 R U R' U' (Rw2 F2 U2 r2 U2 F2 Rw2) U R U' R' y2 and at last the last corner?

I have been checking for parity in all my 5BLD attempts and just done a mental U before memorizing ^^

 

[Zane_C]

About 5BLD, am I correct in this?:

Can anyone write this even simpler? I am sorry but it still is not very clear to me. Do we do pairs of corners untill last corner, then midges with Ui F2 U m2 Ui F2 U, then y2 R U R' U' (Rw2 F2 U2 r2 U2 F2 Rw2) U R U' R' y2 and at last the last corner?

I have been checking for parity in all my 5BLD attempts and just done a mental U before memorizing ^^

It's very unnecessary to have to think about the U center being rotated. Here's the approach that I use:

Memo: xcenters > +centers > wings > corners > midges.
I use images for everything except for the midges.

Solving:
1. Midges: M2.
These are the last pieces to be memorised so can be executed straight away.
I'm not confident enough with edge comms to use UF as my buffer, however I'm thinking about switching to UF and solving difficult cases with M2 - but instead shooting to DB.

You will first need to make sure the algs you use are center friendly.
Instead of shooting to UF with U2 M' U2 M', use a commutator such as [M D2 M'] U2 [M D2 M'] U2 M2.
Instead of shooting to DB with M U2 M U2, use a commutator such as x [M' U2 M] D2 [M' U2 M] D2 M2 x'. (I know, they're bad.)

You will know if you have parity because you memo the corners before the midges.
If you have parity, solve all the midges except leaving UB and UL swapped.

2. xcenters: Comms.
Solve normally.

3. +centers: Comms.
Solve normally.

4. Wings: r2.
Solve normally.

5. Corners: BH/OP.
On the last corner target you will be forced to use either Y, J or L-perm. This will solve the last corner and also swap the UL and UB midges into their correct positions.
However, the UL and UB wing pairs are swapped in the process. Fix this by: B U' B' (Rw2 F2 U2 r2 U2 F2 Rw2) B U B'.

Feel free to ask if you're still unsure on something.

 

[MrMoney]

Thanks Zane, it is really clear now. When doing midges I always do commutators for the UF and DB cases.

 

[Zane_C]

I'm interested in what approach the 5BLD solvers here use to flip midges. I find that using center friendly conjugates is slow.

 

[aronpm]

I use m' U m' U m' U2 m U m U m U2

Not sure if the correct notation is m or M.

It's center-safe, if your centers are solved. If not, it messes them up.

 

[Gaétan Guimond]

M. In my method I begin to solve the corners so I learned to know that with M fingers work differently.

anti hero anti commercial

 

[Mike Hughey]

I use m' U m' U m' U2 m U m U m U2

Not sure if the correct notation is m or M.

It's center-safe, if your centers are solved. If not, it messes them up.

That's what I use too. Flipping those central edges is always the very last thing I do with a 5x5x5 BLD solve, if I have them.

 

[Zane_C]

Thanks, I think I will start executing all the midges last. That way I don't have to try and think of center-safe comms for every single case, I can just incorporate other algs into it.

 

[cmhardw]

I flip midges as the very first part of the solve (just after moving central centers if the solve was a re-oriented solve). I use commutators and cyclic shifts of commutators (setup moves are never necessary).

To flip UF and UR I would do:
R' E' R2 E2 R' U' R E2 R2 E R U

To flip UL and FR I would do a cyclic shift with:
E' R2 E2 R' U2 R E2 R2 E R U2 R'

It's really the commutator:
R' E' R2 E2 R' U2 R E2 R2 E R U2

but I start at the second move and wrap around. This way setup moves are not necessary.

To flip UL and FL I would start at the third move and wrap around with a cyclic shift:
R2 E2 R' U2 R E2 R2 E R U2 R' E'

For 4 midges flipped I'm still using the mono-flip style alg, but done 4 times. I've been lazy honestly and haven't tried to come up with a shorter mono-flip style alg that actually flips 2 edges at the same time.

I would say there is no reason to hang onto the midge flip information until the end of the solve. Sure you all are using faster algs than I am (because I need algs that are fully supercube safe). But at the same time, for any 2 edges flipped case I can solve it in 12 moves. I also can memo the flips visually and do them immediately after putting on the blindfold. That's just my opinion of course. I prefer to use longer moves for the saving in memo time. I feel the net time loss/gain is actually a time gain this way.